WO2024001169A1

WO2024001169A1 - Pedestrian minor-collision identification method and system in low-speed scenario

Info

Publication number: WO2024001169A1
Application number: PCT/CN2023/072689
Authority: WO
Inventors: 丁明慧
Original assignee: 重庆长安汽车股份有限公司
Priority date: 2022-06-30
Filing date: 2023-01-17
Publication date: 2024-01-04
Also published as: CN115056773B; MX2024001538A; EP4365046A1; CN115056773A

Abstract

A pedestrian minor-collision identification method and system in a low-speed scenario. The method comprises: acquiring collision signal sensing sources, involving: 1) calculating a collision probability at a passable area point; 2) calculating a pedestrian target collision probability; and 3) calculating collision probabilities of acceleration sensors; on the basis of signals from the sensing sources and by means of a collision identification system, comprehensively determining whether a pedestrian outside a vehicle has collided with the vehicle at present; and if no collision occurs, the system working normally, and when it is detected that a collision occurs in a corresponding area, the system needing to take a corresponding collision coping strategy, such that the vehicle can immediately identify a minor collision after the collision occurs, and can take a corresponding system strategy at a vehicle end, thereby preventing a pedestrian or animal, which has been hit by the vehicle, outside the vehicle from being run over once again.

Description

A method and system for pedestrian micro-collision recognition in low-speed scenarios

Technical field

The intelligent driving assistance technology of the present invention specifically relates to a pedestrian micro-collision recognition method and system in a low-speed scene.

Background technique

The remote valet parking APA (Automatic Parking Assist) system uses multi-sensor (ultrasonic, millimeter wave, camera, lidar, etc.) fusion detection technology to realize last-mile valet parking, remote valet parking, remote car movement, L4 level driverless functions in limited areas such as one-click summoning. The usage scenario of the remote valet parking APA system is non-public roads in ground or underground parking lots, and its working speed range is 0~15Km/h. Since the remote valet parking APA system is positioned at L4 level, that is, there is no driver in the car and no user remote supervision is required, the system's "observation" ability during the working process completely relies on the detection of the sensor, which due to its own physical Due to the limitations of factors and environmental factors, its perception ability is greatly different from that of humans. It is difficult to identify pedestrians in certain scenarios, and it is easy for pedestrians to be missed and cause collisions.

ISO22737, the world's first technical standard for specific L4 autonomous driving systems, puts forward clear requirements for the protection of pedestrians outside the vehicle during low-speed autonomous driving (LSAD) on predefined routes operating at speeds below 32km/h. This includes collision avoidance and testing of pedestrians outside the vehicle, as well as interaction with the dispatcher after danger occurs. Therefore, identifying collisions with pedestrians outside the vehicle and performing emergency stops, recording collision data, and interacting with the dispatcher are of great significance to the protection of pedestrians outside the vehicle. .

On currently mass-produced vehicles, acceleration sensors or pressure tube sensors are used to identify pedestrian collisions when the vehicle speed is above 25KM/h. Due to the respective physical structures, installation locations and detection principles of the acceleration sensors and pressure tube sensors, Restrictions; however, it is difficult to identify pedestrian collisions with vehicle speeds below 25KM/h, pedestrian collisions behind the vehicle, and pedestrian scratches on both sides of the vehicle. After actual vehicle testing, when the vehicle speed range is 0~15Km/h, the accuracy of using only the acceleration sensor to identify pedestrian micro-collision is approximately 50~75%. In low-speed scenarios, due to the low speed of the vehicle and the relatively soft body of the pedestrian (non-rigid collision), the damage caused by the collision itself is not great. However, because it is difficult to detect and identify the micro-collision using the original collision identification strategy, it is easy to cause a secondary collision. There is a risk of crushing, and in this low-speed scenario, the damage caused by secondary crushing is far more serious than the collision itself.

Aiming at the importance of pedestrian collision recognition in low-speed scenarios, those skilled in the art have conducted research on pedestrian collision detection devices/sensors. For example, "Vehicle Bumper Structure with Pedestrian Collision Detection Sensor" disclosed in CN201580007859.8 (the applicant is Toyota Motor Corporation) provides a bumper structure that can detect a collision with a collision object at the corner of the vehicle based on a collision detection sensor. Bumper structure for vehicles. The "pedestrian collision protection trigger device, pedestrian collision protection device and automobile" disclosed in CN201822036612.2 (the applicant is Zhejiang Geely Automobile Research Institute Co., Ltd.; Zhejiang Geely Holding Group Co., Ltd.) provides a pedestrian collision protection trigger device including a mounting bracket and multiple A plurality of sensors are arranged at intervals on the mounting bracket. The mounting bracket is fixedly connected to the front cross member, and the mounting bracket is located between the front bumper and the front cross beam. The mounting bracket is a deformable structure, and the sensor is an angle sensor or an acceleration sensor. Effectively identify pedestrian collisions. The above-mentioned technology has the problem of high cost because it requires rearrangement or replacement of hardware. Another example is the "Calculation method and safety evaluation system of vehicle-pedestrian collision risk domain" disclosed in CN202010344141.0. The applicant is Tsinghua University, by detecting and outputting vehicle information and pedestrian information, determines whether pedestrians have noticed the vehicle, and determines the collision risk area between vehicles and pedestrians based on the hypothesis of whether pedestrians take active avoidance behavior and whether vehicles take immediate reaction actions, which can effectively improve Pedestrian safety and vehicle driving comfort during the interaction between vehicles and pedestrians. However, detection of targets and reasonable prediction of behavior is only performed before a collision occurs, but detection of pedestrian collisions is not completed.

Contents of the invention

In view of the above-mentioned deficiencies in the existing technology, the purpose of the present invention is to provide a method and system for pedestrian micro-collision recognition in low-speed scenarios, so as to solve the problem that existing collision sensors (acceleration sensors, pressure tube sensors) are difficult to identify vehicles in low-speed driving scenarios. The problem of pedestrian micro-collision is that it is impossible to prevent the vehicle from being crushed twice by a colliding pedestrian or animal in time.

To achieve the above objectives, the present invention adopts the following technical solutions:

A pedestrian micro-collision identification method in a low-speed scene is characterized by including the following steps:

S1. Obtain the collision signal sensing source, including: 1) Calculate the collision probability of the passable area point; 2) Calculate the collision probability of the pedestrian target; 3) Calculate the collision probability of the acceleration sensor;

S2. Based on the signal from the sensing source, use the collision recognition system to comprehensively determine whether a collision with a pedestrian outside the vehicle has occurred;

S3. If no collision occurs, the system works normally; when a collision is detected in the corresponding area, the system needs to adopt corresponding collision response strategies.

Further, in S1, the calculation of the point collision probability of the passable area is based on the vehicle's external sensors (front-view camera, peripheral-view camera, surround-view camera, forward millimeter-wave radar, angular millimeter-wave radar, ultrasonic radar, lidar etc.) respectively detected passable area points; among them, the surround-view camera and ultrasonic radar are necessary sensors for the implementation of the present invention, and the existence of other sensors will make the detection results more reliable.

Each passable area point obtained by multi-sensor information fusion has its own corresponding passable area collision risk. This attribute value will be initialized to 0 when the passable area point is generated; a passable area collision risk of 0 means no collision Risk, 1 means there is a risk of collision; at the same time, the collision risk degree of the passable area of three consecutive passable area points is 1, which means there is a risk of collision, then the output signal: the collision probability of the passable area point is 1, which means high probability, safe The traffic area point collision area is the area where the corresponding traffic area point with collision risk exists.

Further, in S2, the output signal for comprehensively judging whether there is a collision includes: the micro-collision detection status after fusion, whose value is: 0 represents no collision; 1 represents collision; the collision probability of the target after fusion, whose value range is 0 ~100%; the value of the micro-collision area after fusion is: 1 represents no collision, 2 represents the front collision area, 3 represents the rear collision area, 4 represents the left collision area, and 5 represents the right collision area.

Further, in S3, the output of the collision response strategy is: the micro-collision detection status after fusion is 1, which represents a collision; the micro-collision area after fusion is a specific collision area. The system collision response strategy in step S3 is that when the micro-collision detection status is 1 after fusion, it represents a collision, that is, it is determined that the vehicle collides with a pedestrian target outside the vehicle.

The present invention also provides a pedestrian micro-collision recognition system in a low-speed scenario, which is characterized in that it includes an external sensor of the vehicle, an acceleration sensor and a processor. The processor executes the above-mentioned pedestrian micro-collision recognition method in a low-speed scenario.

Compared with the existing technology, the present invention has the following beneficial effects:

1. This invention is based on the existing acceleration sensor of the vehicle, and through the development of back-end strategies, it effectively solves the problem that traditional collision sensors (acceleration sensor, pressure tube sensor) are difficult to identify pedestrian micro-collision in low-speed vehicle driving scenarios, and accurately It can accurately identify collisions of adults, children, pets, etc. at low speeds, so that the vehicle can immediately identify micro-collision after a collision, and adopt corresponding system strategies on the vehicle side to avoid secondary crushing by pedestrians or animals outside the vehicle. Ensure the safety of life outside the vehicle. Specifically, it involves the rapid identification strategy after a micro-collision between the vehicle and pedestrians outside the vehicle (including children with a height of over 80cm and a weight of over 10kg) when the vehicle is driving at low speed and reversing. This system can effectively avoid secondary crushing after a pedestrian collision. pressure.

2. Without adding other sensors and other hardware facilities, the present invention effectively solves the problem of identifying micro-collision between the front collision area, rear collision area, left collision area and right collision area of the vehicle and pedestrians in low-speed driving scenarios, and avoids pedestrian collisions. The second crushing effectively protects the life safety of pedestrians outside the vehicle; it also greatly saves costs and reduces the time for hardware development and matching.

3. The present invention effectively solves the problem of the installation position of traditional collision sensors, which results in the inability to fully identify micro-collision of pedestrians in the front, rear, left and right of the vehicle, and improves the safety protection capability of the vehicle.

Description of drawings

Figure 1 is a schematic diagram of the protection range of pedestrian micro-collision recognition in low-speed scenarios.

Figure 2 is a flow chart of a pedestrian collision recognition system in a low-speed scenario according to the present invention.

Figure 3 is a flow chart of collision risk in a passable area in a low-speed scenario according to the present invention.

Figure 4 is a flow chart of point collision probability in a passable area in a low-speed scenario according to the present invention.

Figure 5 is a flow chart of target collision risk in a low-speed scenario according to the present invention.

Figure 6 is a flow chart of pedestrian target collision probability in a low-speed scenario according to the present invention.

Detailed ways

In order to enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be further described in detail below with reference to specific embodiments and drawings, but the implementation of the present invention is not limited thereto.

In the present invention, the physical meanings of the parameters and signals involved are as follows:

Table 1 Process parameters of a pedestrian micro-collision recognition system in low-speed scenarios

Table 2 Output signals of a pedestrian micro-collision recognition system in a low-speed scenario

Table 3 Comparison table of Chinese and English names of the present invention

As shown in Table 1, it is the process parameters of the pedestrian micro-collision recognition system in low-speed scenarios involved in the present invention. In order to meet the requirements of the pedestrian micro-collision recognition system configured with different vehicle models, the process parameters need to be based on actual vehicles configured with different vehicle models. Test data are individually calibrated.

The process parameter assignment in the present invention is only a calibration value for the currently adapted vehicle model. The process parameters include: the maximum value of the longitudinal distance of the pedestrian collision area K_Pflogcollision, the absolute value of the lateral distance of the pedestrian collision area K_Phorlision, the threshold of pedestrian collision acceleration K_Ahorlision, the weight of the point collision probability of the passable area K_a, the weight of the pedestrian target collision probability K_b, The weight K_c of the collision probability of the acceleration sensor and the collision threshold line K_Cp.

As shown in Table 2, it is the pedestrian micro-collision recognition system in low-speed scenarios involved in the present invention. Its output signal includes a module that calculates the point collision probability of the passable area and outputs a total of 3 signals, which are respectively the collision risk of the passable area, which is A value of 0 represents no collision risk; 1 represents a collision risk; the value of the passable area point collision area is 1 represents no collision; 2 represents the front collision area; 3 represents the rear collision area; 4 represents the left collision area; 5 represents the right Collision zone; the collision probability of a point in the passable area is 0, which represents no collision; 1, which represents a high probability.

The module for calculating the pedestrian target collision probability outputs a total of 3 signals, which are the pedestrian target collision risk degree, and its value range is from 0 to 10 levels; the pedestrian target collision probability, with a value of 0 representing no collision; 1 representing medium probability; 2 Represents high probability; pedestrian target collision area, with a value of 1 representing collision; 2 representing front collision area; 3 representing rear collision area; 4 representing left collision area; 5 represents the right collision area.

The collision probability calculation module of the acceleration sensor outputs a total of 1 signal, which is the collision probability of the acceleration sensor. Its value is 1, which represents low probability; 2, which represents medium probability; and 3, which represents high probability.

The invention finally outputs 3 signals, which are the micro-collision detection status after fusion, with a value of 0 representing no collision; 1 representing collision; the collision probability of the target after fusion, which ranges from 0 to 100%; the micro-collision area after fusion Its value is 1 for no collision; 2 for front collision area; 3 for rear collision area; 4 for left collision area; 5 for right collision area.

Children refer to people who are over 90cm tall and weigh over 10Kg;

Pets refer to animals weighing more than 5Kg.

Refer to Figure 1, a schematic diagram of the pedestrian collision recognition range in low-speed scenarios, including 4 areas: front collision area, rear collision area, left collision area, and right collision area.

The establishment of the coordinate system of the vehicle: the center of the rear axis is the origin of the coordinates, which conforms to the right-hand rule (the horizontal direction is the The vehicle width is W, the longitudinal distance between the center of the rear axle and the front bumper is L1, and the longitudinal distance between the center of the rear axle and the rear bumper is L2. The front collision area refers to the T-shaped area in front of the vehicle under the coordinate system of the vehicle. The four corner points of the T-shaped area are (-W/2,L1), (W/2,L1), (W/2+ 2,L1+3), (-W/2-2,L1+3). The rear collision area refers to the T-shaped area behind the vehicle in the coordinate system of the vehicle. The four corner points of the T-shaped area are (-W/2,-L2), (W/2,-L2), (W/ 2+2,-L2-3), (-W/2-2,-L2-3). The left collision area refers to the T-shaped area to the left of the vehicle in the coordinate system of the vehicle. The four corner points of the T-shaped area are (W/2, L1), (W/2,-L2), (W/2 +2,-L2-3), (W/2+2,L1+3). The right collision area refers to the T-shaped area to the right of the vehicle in the coordinate system of the vehicle. The four corner points of the T-shaped area are (-W/2, L1), (-W/2,-L2), (- W/2-2,-L2-3), (-W/2-2,L1+3). The collision identification system is planned to use the fusion results of surround view and ultrasonic waves, which can effectively protect the 3 meters in front of the car's front bumper, 3 meters behind the rear bumper, and 2 meters to the left and right of the outer edge of the car.

As shown in Figure 2, a pedestrian micro-collision identification method in low-speed scenes includes the following steps:

S1. Obtain the collision signal sensing source: 1) Calculate the collision probability of the passable area point; 2) Calculate the collision probability of the pedestrian target; 3) Calculate the collision probability of the acceleration sensor;

S2. Based on the signal of the sensing source, the collision recognition system is used to comprehensively determine whether a collision with a pedestrian outside the vehicle has occurred. The output signal includes: the micro-collision detection status after fusion, and its value is: 0 represents no collision; 1 represents Collision; the collision probability of the target after fusion, its value range is 0 ~ 100%; the micro-collision area after fusion, its value is: 1 represents no collision, 2 represents the front collision area, 3 represents the rear collision area, 4 represents the left Collision area, 5 represents the right collision area;

S3. If no collision occurs, the system works normally; when a collision is detected in the corresponding area, the system needs to adopt a corresponding collision response strategy and output at the same time: after fusion, the micro-collision detection status is 1, which represents a collision; after fusion, the micro-collision area is Specific collision area.

Among them, 1) Calculate the collision probability of the passable area point; 2) Calculate the collision probability of the pedestrian target; 3) Calculate the collision probability of the acceleration sensor; and the specific content of the collision identification strategy is as follows:

1. Calculate the probability of point collision in the passable area

In the present invention, the external sensors of the vehicle include passable area points detected by forward-looking cameras, peripheral-view cameras, surround-view cameras, forward millimeter-wave radars, angular millimeter-wave radars, ultrasonic radars, laser radars, etc., among which surround-view cameras Cameras and ultrasonic radar are necessary sensors for the implementation of the present invention, and the presence of other sensors will make the detection results more reliable. Among them, the generation of passable area points by sensors and the fusion process of passable area points are all existing technologies, such as CN202011306718.5, CN201810524658.0, CN201910007212.5, etc., which are all involved and are not within the protection scope of the present invention. Calculation and judgment are performed based on the passable area points after fusion of each sensor.

Each passable area point obtained by multi-sensor information fusion has its own corresponding passable area collision risk. This attribute value will be initialized to 0 when the passable area point is generated; a passable area collision risk of 0 means no collision Risk, a value of 1 represents a risk of collision. At the same time, if the passable area collision risk of three passable area points is 1, it means there is a collision risk, then the output signal is: the passable area point collision probability is 1, which means it is high, and the passable area point collision area is the corresponding collision risk. The existence area of the passable area point. The output signals of the passable area point collision probability module include: a passable area collision risk value of 0 represents no collision risk; 1 represents a collision risk; a passable area point collision area, a value of 1 represents no collision; 2 represents the front collision area; 3 represents the rear collision area; 4 represents the left collision area; 5 represents the right collision area; point collision probability in the passable area, 0 represents none; 1 represents high probability.

As shown in Figure 3, the calculation process of collision risk in accessible areas is:

(1) Traverse all passable area points obtained by fusion of multi-sensor information;

(2) Judgment: Current frame detection of passable area points;

a) is not true, then the collision risk in the accessible area is 0;

b) is true, then jump to (3);

(3) Judgment: The passable area point has not been detected in the first three cycles;

a) is not true, then the collision risk in the accessible area is 0;

b) is true, then jump to (4);

(4) Judgment: |Longitudinal distance of passable area points|≤K_Pflogcollision;

a) is not true, then the collision risk in the accessible area is 0;

b) is true, then jump to (5);

(5) Judgment: |Transverse distance of passable area points|≤K_Phorlision;

a) is not true, then the collision risk in the accessible area is 0;

b) is true, then jump to (6);

(6) The collision risk in the accessible area is 1;

Calculate the collision probability of the passable area point in the current period based on the "passable area collision risk" attribute value of each passable area point.

As shown in Figure 4, the calculation process of point collision probability in the passable area is:

(1) The collision risk of the accessible area traversing all accessible area points;

(2) Judgment: All passable area points have been traversed;

a) is true, then jump to (6);

b) is not true, then jump to (3);

(3) Judgment: The collision risk in the accessible area is 0;

a) is true, then jump to (1);

b) is not established, then jump to (4);

(4) The number j of high collision risk passable area points: j=j+1;

(5) Judgment: The number of high collision risk passable area points j ≥ 3;

a) is established, the collision probability of the passable area point is high probability. Based on the position of the passable area point, the collision area of the passable area point is output: 1 represents no collision; 2 represents the front collision area; 3 represents the rear collision area; 4 Represents the left collision area; 5 represents the right collision area, jump (6);

b) is not established, then jump to (1);

(6)End;

2. Calculate the collision probability of pedestrian targets

Each target obtained by multi-sensor information fusion has its own corresponding pedestrian collision risk. This attribute value will be initialized to 0 when the target is generated. The maximum value of this attribute value is 10. During the calculation process, if the attribute value is greater than 10 Then let its value be 10, and update this attribute every cycle. Among them, the target generated by each sensor and the target fusion process are not within the scope of the present invention. The present invention performs calculation and judgment based on the target results after fusion of each sensor.

The output signals of the module for calculating the collision probability of pedestrian targets include: the collision risk of pedestrian targets ranges from 0 to 10; the collision probability of pedestrian targets, with a value of 0 representing no collision; 1 representing medium probability; 2 representing high probability; Target collision area, the value of which is 1 represents no collision; 2 represents the front collision area; 3 represents the rear collision area; 4 represents the left collision area; 5 represents the right collision area.

As shown in Figure 5, the calculation process of pedestrian target collision risk is:

(1) Traverse all targets obtained by fusion of multi-sensor information;

(2) Judgment: The target type is: pedestrian, animal or other;

a) is not true, then the pedestrian target collision risk is: i=i;

b) is true, then jump to (3);

(3) Judgment: The pedestrian is not detected in the current frame;

a) is not true, then the pedestrian target collision risk is: i=i;

b) is true, then jump to (4);

(4) Judgment: The effective pedestrian tracking period is ≥ 3 frames;

a) is not true, then the pedestrian target collision risk is: i=i;

b) is true, then jump to (5);

(5) Judgment: |Pedestrian longitudinal distance|≤K_Pflogcollision;

a) is not true, then the pedestrian target collision risk is: i=i;

b) is true, then jump to (6);

(6) Judgment: |Pedestrian lateral distance|≤K_Phorlision;

a) is not true, then the pedestrian target collision risk is: i=i;

b) is true, then jump to (7);

(7) Judgment: Vehicle longitudinal acceleration in the first 100ms ≤ K_Ahorlision;

a) is established, then the pedestrian target collision risk degree: i=i;

b) is not established, then jump to (8);

(8) Pedestrian target collision risk: i=i+1;

Calculate the pedestrian target collision probability in the current cycle based on the pedestrian target collision risk of each target. As shown in Figure 6, the calculation process of pedestrian target collision probability is:

(1) Pedestrian target collision risk across all targets;

(2) Judgment: The target has been completely traversed;

a) is true, then jump to (5);

b) is not true, then jump to (3);

(3) Judgment: Pedestrian target collision risk: i≥3;

a) is established, then the pedestrian target collision probability is a high probability. Based on the pedestrian's position, the pedestrian target collision is output. Area: 1 represents no collision; 2 represents front collision area; 3 represents rear collision area; 4 represents left collision area; 5 represents right collision area, jump (5);

b) is not established, then jump to (4);

(4) Judgment: Pedestrian target collision risk: 3≥i≥0;

a) is established, then the pedestrian target collision probability is a medium probability. Based on the pedestrian's position, the pedestrian target collision area is output: 1 represents no collision; 2 represents the front collision area; 3 represents the rear collision area; 4 represents the left collision area; 5 Represents the right collision area, jump(5);

b) is not established, then jump to (1);

(5)End.

3. Calculate the collision probability of the acceleration sensor

The vehicle is equipped with 3 acceleration sensors. The collision of a single acceleration sensor needs to be experimentally calibrated based on the vehicle type in which the sensor is installed. According to the vehicle type where the sensor is installed, the experimental calibration is passed in the early stage and the following calibration results are planned to be used: The peak acceleration value of the collision waveform of a single acceleration sensor is less than 2g, the collision probability of the sensor is considered to be low probability; when the peak acceleration value of the collision waveform of a single acceleration sensor is greater than 2g and less than 5g, the collision probability of the sensor is considered to be medium probability; the peak acceleration value of the collision waveform of a single acceleration sensor is greater than At 5g, the collision probability of this sensor is considered to be high.

The output signal of the collision probability calculation module of the acceleration sensor includes: the collision probability of the acceleration sensor, with a value of 1 representing low probability; 2 representing medium probability; and 3 representing high probability.

As shown in Table 3, the following judgment is made based on the values of the three acceleration sensors. This judgment believes that the collision detection probability of the three acceleration sensors has nothing to do with the installation position of the sensors. Therefore, there is no difference in the order and position of the three sensors in the table. The collision probability of the acceleration sensor represents the assessment of the current possibility of a collision from the perspective of the acceleration sensor. Its value is divided into three levels: 1 represents low probability, 2 represents medium probability, and 3 represents high probability.

1) The collision probabilities of the three acceleration sensors are all low probability, then the collision probability is: low probability.

2) The collision probabilities of the three acceleration sensors are all medium probability, then the collision probability is: medium probability.

3) The collision probabilities of the three acceleration sensors are all high probability, then the collision probability is: high probability.

4) The collision probabilities of the three acceleration sensors are high probability, medium probability, and low probability respectively. The collision probability is: medium probability.

5) The collision probability of two acceleration sensors is low probability, and the collision probability of another acceleration sensor is medium probability, then the collision probability is: low probability.

6) The collision probability of two acceleration sensors is low probability, and the collision probability of another acceleration sensor is high probability, then the collision probability is: medium probability.

7) The collision probability of two acceleration sensors is medium probability, and the collision probability of another acceleration sensor is low probability, then the collision probability is: low probability.

8) The collision probability of two acceleration sensors is medium probability, and the collision probability of another acceleration sensor is high probability, then the collision probability is: medium probability.

9) The collision probability of two acceleration sensors is high probability, and the collision probability of another acceleration sensor is low probability, then the collision probability is: medium probability.

10) The collision probability of two acceleration sensors is high probability, and the collision probability of another acceleration sensor is medium probability, then the collision probability is: high probability.

Table 3 Collision probability calculation strategy of acceleration sensor

4. Collision identification strategy

Based on the passable area point collision probability, pedestrian target collision probability, and acceleration sensor collision probability obtained by the above calculation method, corresponding weights are assigned to the three signal sources according to the accuracy of the collision signal source, in order K_a, K_b, K_c, The weights need to be calibrated based on actual vehicle testing. The weight values here are temporarily assigned to 0.5, 0.2, and 0.3 based on experience. At the same time, the corresponding collision probabilities under the results are set according to the detection results of the three signal sources, as shown in Table 4. Show.

The output signals of the collision recognition strategy module include: micro-collision detection status after fusion, with a value of 0 representing no collision; 1 representing collision; collision probability of the target after fusion, with a value ranging from 0 to 100%; micro-collision after fusion Area, the value of which is 1 represents no collision; 2 represents the front collision area; 3 represents the rear collision area; 4 represents the left collision area; 5 represents the right collision area.

Based on the above analysis, the collision probability of the fused target is calculated as: C=K_a*Ai+K_b*Bi+K_c*Ci; where C>K_Cp is determined to be a current collision, and K_Cp is the collision threshold line that needs to be calibrated according to actual vehicle testing. , the collision threshold line here is temporarily assigned a value of 50% based on experience. Output signal: After fusion, the micro-collision detection status is 1, which represents collision, and the micro-collision area after fusion is the collision area of the passable area point.

When the signal passable area point collision area is the pedestrian target collision area and is not zero, the collision probability of the target after fusion is C=1.2C; where C≤100%; that is, when C>100%, the mandatory order C=100% .

In the embodiment, the collision probability Ai of the passable area point at a certain moment is a high probability, the collision area of the passable area point is the front collision area, the pedestrian target collision probability Bi is a medium probability, the pedestrian target collision area is the front collision area, and the acceleration sensor The collision probability Ci is a low probability, then the collision probability of the target after fusion is calculated as: C=0.5*100%+0.2*50%+0.3*0%=60%; because the collision area of the passable area point is the pedestrian target collision area , which is the front collision area, C=1.2C=72%, the collision probability C of the target after fusion is 72%>K_Cp, and the micro-collision detection state after fusion is determined to be collision.

Table 4 Collision probabilities of three signal sources

5. System collision response strategies

When the micro-collision detection status after fusion is 1, it indicates a collision, that is, it is judged that the vehicle collided with a pedestrian target outside the vehicle. The system adopts the corresponding collision response strategy as follows:

1) The vehicle stops immediately at maximum deceleration;

2) Vehicles are prohibited from driving during the current ignition cycle;

3) Double flashes when the vehicle applies the handbrake, locks, raises the window, turns off the engine, and opens the vehicle;

4) Prompt the user that a collision is currently occurring, prompt the user that the current collision area is: the post-fusion micro-collision area, and request the user to take over;

5) Prompt that a collision is currently occurring at the parking lot where the vehicle is located, and prompt that the current collision area at the site is: the post-fusion micro-collision area, and request the site dispatcher to handle it in a timely manner;

6) Vehicles with V2X functions need to remind surrounding vehicles through V2X: This vehicle has collided, please pay attention to avoid it safely.

This invention is aimed at the limitations of sensor perception and the inability of traditional collision sensors (acceleration sensors) to accurately detect micro-collision in all directions in front, rear, left and right micro-collision scenarios when vehicles are traveling at low speeds (0-15Km/h). problem, without adding additional sensors and hardware facilities, accurately identify collisions of adults, children, pets, etc. at low speeds, so that the vehicle can immediately identify micro-collision after a collision, and adopt corresponding system strategies on the vehicle side , to prevent secondary crushing by pedestrians or animals outside the vehicle and to ensure the safety of life outside the vehicle.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the technical solutions. Those of ordinary skill in the art should understand that those technical solutions of the present invention can be modified or equivalently substituted without departing from the present technology. The purpose and scope of the solution should be covered by the claims of the present invention.

Claims

A pedestrian micro-collision identification method in a low-speed scene is characterized by including the following steps:

S1. Obtain the collision signal sensing source, including: 1) Calculate the collision probability of the passable area point; 2) Calculate the collision probability of the pedestrian target; 3) Calculate the collision probability of the acceleration sensor;

S2. Based on the signal from the sensing source, use the collision recognition system to comprehensively determine whether a collision with a pedestrian outside the vehicle has occurred;

S3. If no collision occurs, the system works normally; when a collision is detected in the corresponding area, the system needs to adopt corresponding collision response strategies.
The pedestrian micro-collision identification method in a low-speed scene according to claim 1, characterized in that, in S1, calculating the passable area point collision probability is based on the vehicle's external sensors, that is, the passable area detected by the surround-view camera and ultrasonic radar. Area points, calculate and judge the passable area points after fusion of each sensor;

Each passable area point obtained by multi-sensor information fusion has its own corresponding passable area collision risk. This attribute value will be initialized to 0 when the passable area point is generated. A passable area collision risk of 0 means no collision. Risk, 1 means there is a risk of collision; at the same time, the collision risk degree of the passable area of three consecutive passable area points is 1, which means there is a risk of collision, then the output signal: the collision probability of the passable area point is 1, which means high probability, safe The traffic area point collision area is the area where the corresponding traffic area point with collision risk exists.
The pedestrian micro-collision identification method in a low-speed scene according to claim 1, characterized in that in S1, 2) calculating the pedestrian target collision probability, including each target obtained by multi-sensor information fusion having its own corresponding pedestrian collision Risk degree. This attribute value will be initialized to 0 when the target is generated. The maximum value of this attribute value is 10. During the calculation process, if the attribute value is greater than 10, the value will be 10. This attribute will be updated every cycle. .
The pedestrian micro-collision identification method in a low-speed scene according to claim 1, characterized in that in S1, the collision probability of the acceleration sensor is calculated in 3), the vehicle is equipped with three acceleration sensors, and the peak acceleration of the collision waveform of a single acceleration sensor When the value is less than 2g, the collision probability of the sensor is considered to be low probability; when the peak acceleration value of the collision waveform of a single acceleration sensor is greater than 2g and less than 5g, the collision probability of the acceleration sensor is considered to be medium; the peak acceleration value of the collision waveform of a single acceleration sensor When it is greater than 5g, the collision probability of the sensor is considered to be high.
The pedestrian micro-collision identification method in a low-speed scene according to claim 1, characterized in that, in the S2, the output signal for comprehensively determining whether there is a collision includes: the micro-collision detection status after fusion, and its value is: 0 represents no collision. ; 1 represents collision; the collision probability of the target after fusion, its value range is 0 ~ 100%; the value of the micro-collision area after fusion is: 1 represents no collision, 2 represents the front collision area, 3 represents the rear collision area, 4 represents the left collision area and 5 represents the right collision area.
The pedestrian micro-collision identification method in a low-speed scene according to claim 1, characterized in that, in the S3, the collision response strategy output is: the micro-collision detection status after fusion, 1 represents collision; the micro-collision area after fusion, is Specific collision area.
The pedestrian micro-collision identification method in a low-speed scene according to claim 2, wherein the vehicle external sensor further includes a forward-looking camera, a peripheral-view camera, a forward millimeter wave radar, an angular millimeter wave radar, and a laser radar.
The pedestrian micro-collision identification method in a low-speed scene according to claim 2, characterized in that the output signal of the module for calculating the point collision probability of the passable area includes: a collision risk degree of the passable area, with a value of 0 representing no collision risk; 1 represents the risk of collision; the passable area points to the collision area, and its value is 1, which represents no collision; 2 represents the front collision area; 3 represents the rear collision area; 4 represents the left collision area; 5 represents the right collision area; the passable area point Collision probability, 0 represents none, 1 represents high probability.
The pedestrian micro-collision identification method in a low-speed scene according to claim 3, characterized in that the output signal of the pedestrian target collision probability calculation module includes: a pedestrian target collision risk degree, the value range of which is 0 to 10 levels; Collision probability, the value of which is 0 represents no collision; 1 represents medium probability; 2 represents high probability; pedestrian target collision area, whose value is 1 represents no collision; 2 represents the front collision area; 3 represents the rear collision area; 4 represents Left collision area; 5 represents the right collision area.
The pedestrian micro-collision identification method in a low-speed scene according to claim 4, characterized in that the output signal of the module for calculating the collision probability of the acceleration sensor includes: the collision probability of the acceleration sensor, with a value of 1 representing low probability; 2 representing medium Probability; 3 represents high probability.
The pedestrian micro-collision identification method in a low-speed scene according to claim 4, characterized in that the following judgment is made based on the values of the three acceleration sensors. This judgment believes that the collision detection probability of the three acceleration sensors has nothing to do with the installation position of the sensors, so There is no difference in order or position between the three sensors; the collision probability of the acceleration sensor represents the assessment of the current possibility of a collision from the perspective of the acceleration sensor. Its value is divided into three levels: 1 represents low probability, 2 represents medium probability, and 3 represents high probability. ;

The collision probabilities of the three acceleration sensors are all low probability, then the collision probability is: low probability;

The collision probabilities of the three acceleration sensors are all medium probability, then the collision probability is: medium probability;

The collision probabilities of the three acceleration sensors are all high probability, then the collision probability is: high probability;

The collision probabilities of the three acceleration sensors are high probability, medium probability, and low probability respectively. The collision probability is: medium probability;

The collision probability of two acceleration sensors is low probability, and the collision probability of another acceleration sensor is medium probability, then the collision probability is: low probability;

The collision probability of two acceleration sensors is low probability, and the collision probability of another acceleration sensor is high probability, then the collision probability is: medium probability;

The collision probability of two acceleration sensors is medium probability, and the collision probability of another acceleration sensor is low probability, then the collision probability is: low probability;

The collision probability of two acceleration sensors is medium probability, and the collision probability of another acceleration sensor is high probability, then the collision probability is: medium probability;

The collision probability of two acceleration sensors is high probability, and the collision probability of another acceleration sensor is low probability, then the collision probability is: medium probability;

The collision probability of two acceleration sensors is high probability, and the collision probability of another acceleration sensor is medium probability, then the collision probability is: high probability.
The pedestrian micro-collision identification method in a low-speed scene according to claim 5, characterized in that the S2 collision identification system comprehensive judgment method includes the passable area point collision probability, pedestrian target collision probability and acceleration obtained based on the above calculation method. The collision probability of the sensor assigns corresponding weights to the three signal sources according to the accuracy of the collision signal source, in order K_a, K_b, K_c. The weights need to be calibrated based on actual vehicle testing;

The output signals of the collision recognition strategy module include: micro-collision detection status after fusion, with a value of 0 representing no collision; 1 representing collision; collision probability of the target after fusion, with a value ranging from 0 to 100%; micro-collision after fusion Area, the value of which is 1 represents no collision; 2 represents the front collision area; 3 represents the rear collision area; 4 represents the left collision area; 5 represents the right collision area.
The pedestrian micro-collision identification method in a low-speed scene according to claim 12, characterized in that the calculation of the collision probability of the fused target is: C=K_a*Ai+K_b*Bi+K_c*Ci;

Among them, when C>K_Cp, it is determined that a collision is currently occurring. K_Cp is the collision threshold line and needs to be calibrated according to the actual vehicle test. The collision threshold line here is temporarily assigned a value of 50% based on experience; the output signal: the micro-collision detection status after fusion is 1. Collision, the micro-collision area after fusion is the collision area of the passable area point.
The pedestrian micro-collision identification method in a low-speed scene according to claim 12 or 13, characterized in that when the signal passable area point collision area is a pedestrian target collision area and is not zero, the collision probability of the fused target is C=1.2 C; where C≤100%; that is, when C>100%, C=100% is mandatory.
The pedestrian micro-collision identification method in a low-speed scene according to claim 6, characterized in that the system collision response strategy in step S3 is that when the micro-collision detection status is 1 after fusion, it represents a collision, that is, it is judged whether the vehicle and the pedestrian outside the vehicle are When a target collides, the system adopts the corresponding collision response strategy as follows:

1) The vehicle stops immediately at maximum deceleration;

2) Vehicles are prohibited from driving during the current ignition cycle;

3) Double flashes when the vehicle applies the handbrake, locks, raises the window, turns off the engine, and opens the vehicle;

4) Prompt the user that a collision is currently occurring, prompt the user that the current collision area is: the post-fusion micro-collision area, and request the user to take over;

5) Prompt that a collision is currently occurring at the parking lot where the vehicle is located, prompt that the current collision area at the site is a post-fusion micro-collision area, and request the site dispatcher to handle it in a timely manner;

6) Vehicles with V2X functions need to remind surrounding vehicles through V2X: This vehicle has collided, please pay attention to avoid it safely.
A pedestrian micro-collision recognition system in a low-speed scene, characterized by including an external sensor of the vehicle, an acceleration sensor and a processor, and the processor executes the pedestrian micro-collision recognition method in any low-speed scene of claims 1 to 15.